search

asgeorges / mapper-classifier

This repository contains the initial PoC software to build a mapper based classifier, which utilizes tools from topological data analysis. The package here does not scale well & is optimally inefficient (you're welcome) - however, work is currently being done to refactor this code into a polished package.
7 stars 1 forks source link

Missing trueexamples_ In_ 10_ 10_ PROJ1_ Nodes.txt #12

Open yanglilongss opened 8 months ago

yanglilongss commented 8 months ago

微信截图_20240119182115 Hello, I found that your code is missing '..' while using it/ Data Temp/trueeexamples In 10 10 PROJ1 The 'nodes. txt' file, after reviewing the code, I did not find the code generated by this file. May I ask where the problem lies? The operating system is Win10 and R is installed

asgeorges commented 8 months ago

Eh - I feel sorry you're going through our code. Not our finest work. Did you get all the environment bugs resolved? I haven't touched this repo in years so the env stuff is super outdated.

The above error to me seems like the pymapper or mnist_mapper_light_template1D isn't outputting the txt files correctly, which also probably means env error somewhere.

Anyways, at this point in time, I imagine there's much better code bases out there to do what we did. Or at the very least, there should be better codebases that do portions of our research, then you can string those together. I remember liking these guys a lot.

Lastly - I recommend you also reach out to Jacek Cyranka who still is activeish in this field of research.

asgeorges commented 8 months ago

If you really want to keep using this codebase, happy to debug w/ you 1-1 or feel free to provide more debugging logs

yanglilongss commented 8 months ago

微信截图_20240120220939 Thank you very much for your reply. The above issue has been resolved, and as you mentioned, it is indeed pymapper and mnist Mapper Light_ The issue with template1D, the entire code has been run smoothly. I have another question that I would like to ask you, is the data in the images folder not updated under the operation in the above image? Or is it necessary to run code generation separately, because I am just starting to get involved in this field, so I am not very familiar with some aspects. Sorry to bother you

asgeorges commented 8 months ago

Hey I can't remember at all and I looked through the code too. You might have to run the 1st code block as well.

Jacek do you remember?

dzako commented 8 months ago

Hello yanglilongss,

I am not sure what you mean, to help you more can you clarify your question ?

Also I have a newer version of the code, we recently figured out that the pipeline can be improved by performing SVM as the classifier on the binarized Mapper features. It also utilizes the giotto TDA python package for computing the Mapper graphs instead of the outdated R code. It was written by a former student of mine, never made it public, but if you are interested in following up this line of research, I can share it for you and do some introduction. In such case please reach me out on e-mail (jcyranka at gmail).

Best, Jacek